Get email delivery of the Cadence blog featured here
How do we achieve highly accurate aging data models for critical circuits in automotive or IoT applications?
IoT device aging isn’t well understood yet, since most of it is still so new. How will the software stand up against tomorrow’s threats? How will it operate when surrounded by tomorrow’s Internet? These questions are critical as early IoT devices push their first five years, but more time is still needed to determine the answers.
For now, though, we can focus on making the hardware as safe as possible.
GLOBALFOUNDRIES builds chips for all sorts of applications, including 5G infrastructure, DAS/FAS, hyper-scale data centers, deep neural networks, IoT devices, infotainment and driving control systems, and factory automation. As such, they’re researching how these new chips built for automotive and IoT applications age. In a recent presentation given at DAC, GLOBALFOUNDRIES outlined how they are analyzing aging with the help of Cadence’s Liberate Characterization Solution.
Chips age in two main ways. As a chip gets older, its BTI (bias temperature instability) increases. This is due to the degradation of silicon-hydrogen bonds in a chip; these bonds break as the chip is heated and stressed. NMOS (N-channel) and PMOS (P-channel) transistors experience time-dependent parametric shifts due to Vgs stress; this also contributes to increased BTI.
The other way is HCI (hot carrier injection) stress. This occurs when an electron gains enough kinetic energy to break through a barrier in the chip, causing it to end up somewhere it’s not supposed to be. Normally, this isn’t a big deal, but if it happens too often, it can cause degradation of a chip’s barriers, causing current leakage.
BTI is frequency independent, while HCI occurs more often at higher frequencies. Likewise, BTI can be repaired, but HCI can’t.
Using Liberate Characterization, GLOBALFOUNDRIES was able to measure the impact of aging on transistors and generate highly accurate aging data models for cells. With Liberate Characterization, they were able to map cell-specific aging and perform experiments that resulted in a 7-11% voltage leakage reduction, and a 20% reduction in pessimism for setup TNS.
The path to better-aging chips is being revealed, and Cadence’s Liberate Characterization Solution is a part of it all. For more information on how GLOBALFOUNDRIES uses Liberate Characterization, take a look at their presentation from DAC 2019.